MR imaging of fat-containing tissues: valuation of two quantitative imaging techniques in comparison with localized proton spectroscopy.

Since lipid protons, consisting mainly of triacylglycerols (TAG), are rather mobile, magnetic resonance imaging (MRI) is ideally suited for the examination of fat-containing tissues such as bone marrow. In contrast to water protons, however, lipid protons are chemically distinct and give rise to at least eight resonance peaks with different T1 and T2 relaxation times in the 1H spectrum. This is why the characterization of fat-containing tissues by quantitative MRI is much more difficult than that of most other tissues. In our study we wanted to examine the accuracy and the potential of a 1H chemical shift imaging (CSI) technique and a multiple spin-echo imaging (MSEI) technique. A stimulated-echo (STEAM) sequence for spatially localized proton spectroscopy was used as the reference method. In the first part of this paper, we describe quantitative imaging experiments which were performed to assess the accuracy of the fat-water separation according to the Dixon method and the bi-exponential decomposition of the MSEI data. For that purpose, we used a two-compartment phantom filled with either an aqueous Gd-DTPA solution and vegetable oil or with two different aqueous Gd-DTPA solutions, respectively. The analysis of the 1H CSI data revealed that the presence of non-methylen protons in neutral fats leads to a slight under-estimation (of about 15%) of the relative fat fraction. The error is described theoretically and verified quantitatively by STEAM measurements. The bi-exponential analysis of the transverse relaxation data, on the other hand, yields reliable T2 values if the relative proton density of both components is higher than 15%. IN the second part of our investigation, the same techniques were applied to acquire data from the subcutaneous fatty tissue, the femoral head, and the lumbar vertebrae of three healthy volunteers. In the bone marrow spectra, only two broad resonances could be resolved; they were superpositions of diverse molecular groups with different T1 and T2 relaxation times. In these cases, localized proton spectroscopy does not provide additional information with respect to 1H CSI. The MSEI data of the three examined fat containing tissue regions were adequately fitted by a bi-exponential function despite the fact that there were much more chemically distinct protons present in fatty tissues.